claude-symphony 0.0.7 → 0.0.8

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-symphony",
-  "version": "0.0.7",
+  "version": "0.0.8",
   "description": "Multi-AI Orchestration Framework - Create new projects with 10-stage development workflow",
   "type": "module",
   "bin": {

package/template/CLAUDE.md

@@ -411,6 +411,36 @@ state/
 
 ---
 
+## Parallel AI Execution Policy
+
+### Overview
+All defined AI models execute the same task in parallel, and each output is synthesized to produce the final result.
+
+### Execution Flow
+1. **Parallel Execution**: Designated models perform the task simultaneously
+2. **Individual Output**: Each model generates `output_modelName.md`
+3. **Synthesis**: ClaudeCode analyzes all outputs and generates `final_output.md`
+
+### Stage Model Assignment
+
+| Stage | Parallel Models | Synthesizer | Output |
+|-------|-----------------|-------------|--------|
+| 01-brainstorm | Gemini + ClaudeCode | ClaudeCode | ideas.md |
+| 03-planning | Gemini + ClaudeCode | ClaudeCode | architecture.md |
+| 04-ui-ux | Gemini + ClaudeCode | ClaudeCode | wireframes.md |
+| 07-refactoring | Codex + ClaudeCode | ClaudeCode | refactoring_report.md |
+| 09-testing | Codex + ClaudeCode | ClaudeCode | tests/ |
+
+> Single-model stages (02, 05, 06, 08, 10) maintain existing behavior
+
+### Synthesis Criteria
+1. **Extract Commonalities**: Prioritize content all models agree on
+2. **Analyze Differences**: Compare pros/cons, select best approach
+3. **Complementary Integration**: Merge unique insights from each model
+4. **Quality Filtering**: Remove low-quality content
+
+---
+
 ## Smart HANDOFF System
 
 > Configuration files: `config/handoff_intelligence.yaml`, `config/memory_integration.yaml`

package/template/config/ai_collaboration.yaml

@@ -2,27 +2,42 @@
 # Multi-AI orchestration modes and collaboration strategies
 
 collaboration_modes:
-  # Parallel Execution: Execute same task with multiple AIs simultaneously
+  # Parallel Execution: Execute same task with multiple AIs simultaneously and synthesize
   parallel_execution:
-    description: "Execute same task with multiple AIs simultaneously and select best result"
+    description: "Execute same task with multiple AIs simultaneously and synthesize"
     enabled: true
 
     stages:
-      - "01-brainstorm"
-      - "02-research"
-
-    models:
-      - "gemini"
-      - "claude"
+      01-brainstorm:
+        parallel_models: [gemini, claudecode]
+        synthesizer: claudecode
+        output: ideas.md
+      03-planning:
+        parallel_models: [gemini, claudecode]
+        synthesizer: claudecode
+        output: architecture.md
+      04-ui-ux:
+        parallel_models: [gemini, claudecode]
+        synthesizer: claudecode
+        output: wireframes.md
+      07-refactoring:
+        parallel_models: [codex, claudecode]
+        synthesizer: claudecode
+        output: refactoring_report.md
+      09-testing:
+        parallel_models: [codex, claudecode]
+        synthesizer: claudecode
+        output: tests/
+
+    synthesis_criteria:
+      - extract_commonalities
+      - analyze_differences
+      - complementary_integration
+      - quality_filtering
 
     merge_strategy:
-      type: "best_of_n"
-      selection_criteria:
-        - quality_score
-        - completeness
-        - creativity
-      require_human_review: true
-      auto_select_threshold: 0.85  # Auto-select when score is 85% or higher
+      type: "synthesis"
+      synthesizer: claudecode
 
   # Sequential Handoff: Sequential transfer between AIs (review chain)
   sequential_handoff:

package/template/stages/01-brainstorm/CLAUDE.md

@@ -38,6 +38,28 @@ Divergent idea generation and requirements exploration stage
 - **Secondary**: ClaudeCode (structuring, feasibility review)
 - **Mode**: YOLO (Container) - autonomous execution mode
 
+## Parallel Execution Protocol
+
+### Models
+- **Primary**: Gemini (creative ideation)
+- **Secondary**: ClaudeCode (structural organization)
+
+### Execution
+1. Gemini: Generate `output_gemini.md`
+2. ClaudeCode: Generate `output_claudecode.md`
+3. ClaudeCode (Synthesizer): Synthesize → `ideas.md`
+
+### Output Files
+- `output_gemini.md` - Gemini results
+- `output_claudecode.md` - ClaudeCode results
+- `ideas.md` - Final synthesized result
+
+### Synthesis Criteria
+1. Extract commonalities first
+2. Analyze differences and select best
+3. Integrate unique insights
+4. Filter low-quality content
+
 ## Goals
 1. Divergent ideation based on project brief
 2. In-depth analysis of user requirements

package/template/stages/03-planning/CLAUDE.md

@@ -33,6 +33,28 @@ System architecture and technology stack decision stage
 - **Primary**: Gemini (architecture design, diagrams)
 - **Mode**: Plan Mode - structured design
 
+## Parallel Execution Protocol
+
+### Models
+- **Primary**: Gemini (architecture design)
+- **Secondary**: ClaudeCode (technical validation)
+
+### Execution
+1. Gemini: Generate `output_gemini.md`
+2. ClaudeCode: Generate `output_claudecode.md`
+3. ClaudeCode (Synthesizer): Synthesize → `architecture.md`
+
+### Output Files
+- `output_gemini.md` - Gemini results
+- `output_claudecode.md` - ClaudeCode results
+- `architecture.md` - Final synthesized result
+
+### Synthesis Criteria
+1. Extract commonalities first
+2. Analyze differences and select best
+3. Integrate unique insights
+4. Filter low-quality content
+
 ## Goals
 1. System architecture design
 2. Final technology stack decision

package/template/stages/04-ui-ux/CLAUDE.md

@@ -33,6 +33,28 @@ User interface and experience design stage
 - **Primary**: Gemini (creative UI design)
 - **Mode**: Plan Mode
 
+## Parallel Execution Protocol
+
+### Models
+- **Primary**: Gemini (creative UI design)
+- **Secondary**: ClaudeCode (UX validation)
+
+### Execution
+1. Gemini: Generate `output_gemini.md`
+2. ClaudeCode: Generate `output_claudecode.md`
+3. ClaudeCode (Synthesizer): Synthesize → `wireframes.md`
+
+### Output Files
+- `output_gemini.md` - Gemini results
+- `output_claudecode.md` - ClaudeCode results
+- `wireframes.md` - Final synthesized result
+
+### Synthesis Criteria
+1. Extract commonalities first
+2. Analyze differences and select best
+3. Integrate unique insights
+4. Filter low-quality content
+
 ## Goals
 1. Wireframe design
 2. User flow definition

package/template/stages/07-refactoring/CLAUDE.md

@@ -38,6 +38,28 @@ Code quality improvement and optimization stage
 - **Secondary**: ClaudeCode (complex refactoring)
 - **Mode**: Deep Dive - in-depth code analysis
 
+## Parallel Execution Protocol
+
+### Models
+- **Primary**: Codex (code optimization)
+- **Secondary**: ClaudeCode (code review)
+
+### Execution
+1. Codex: Generate `output_codex.md`
+2. ClaudeCode: Generate `output_claudecode.md`
+3. ClaudeCode (Synthesizer): Synthesize → `refactoring_report.md`
+
+### Output Files
+- `output_codex.md` - Codex results
+- `output_claudecode.md` - ClaudeCode results
+- `refactoring_report.md` - Final synthesized result
+
+### Synthesis Criteria
+1. Extract commonalities first
+2. Analyze differences and select best
+3. Integrate unique insights
+4. Filter low-quality content
+
 ## Goals
 1. Code quality improvement
 2. Duplication removal

package/template/stages/09-testing/CLAUDE.md

@@ -35,6 +35,28 @@ Test code writing and E2E testing stage
 - **Mode**: Sandbox + Playwright
 - **MCP**: playwright server integration
 
+## Parallel Execution Protocol
+
+### Models
+- **Primary**: Codex (test generation)
+- **Secondary**: ClaudeCode (test validation)
+
+### Execution
+1. Codex: Generate `output_codex.md`
+2. ClaudeCode: Generate `output_claudecode.md`
+3. ClaudeCode (Synthesizer): Synthesize → `tests/`
+
+### Output Files
+- `output_codex.md` - Codex results
+- `output_claudecode.md` - ClaudeCode results
+- `tests/` - Final synthesized test files
+
+### Synthesis Criteria
+1. Extract commonalities first
+2. Analyze differences and select best
+3. Integrate unique insights
+4. Filter low-quality content
+
 ## Goals
 1. Unit test writing
 2. Integration test writing